Record 025 · Models · 2025
50%25.25.L
DeepSeek-R1
DeepSeek
“A reasoning model whose distilled GGUFs made chain-of-thought a local feature.”
R1 and the Qwen/Llama distills put long-think behavior on consumer GPUs. Local boxes started pausing, then answering.
Opened local reasoning models.
- Open reasoning traces
- Distills for consumer GPUs
- Think-then-answer locally
Reasoning weights you can distill onto a 7B.
